WO2005109245A2

WO2005109245A2 - Method and device for the reproduction of information

Info

Publication number: WO2005109245A2
Application number: PCT/EP2005/003959
Authority: WO
Inventors: Merce Serra; Olaf Korte; Alexander Zink
Original assignee: Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V.
Priority date: 2004-04-28
Filing date: 2005-04-14
Publication date: 2005-11-17
Also published as: US20070078730A1; KR100922418B1; DE502005003093D1; CN1950822A; US7937357B2; ATE388447T1; KR20070001249A; CA2564944C; DE102004020878A1; AU2005241161B2; JP2007535048A; HK1095186A1; CN100465958C; JP4630897B2; AU2005241161A1; CA2564944A1; WO2005109245A3; EP1725958B1; EP1725958A2

Abstract

The aim of the invention is to provide an intelligent system for the reproduction of information, which adapts itself to the wishes and requirements of the user in a practically imperceptible manner that involves little or no additional complexity. To achieve this, the information to be reproduced is selected randomly from information or a plurality of information, based on an adaptive probability distribution. In addition, the adaptive probability distribution is adapted in accordance with the detected user behaviour for the reproduction of the selected information. The behaviour of the user is consequently used as a controlled variable for the adaptation of the adaptive probability distribution, the latter defining the probability among the reproducible information or information objects in which the reproducible information is provided that a specific piece of information, contained in an information object, will be selected to be included in the next piece of information to be reproduced in the subsequent random selection.

Description

Method and device for information reproduction

description

The present invention relates to information reproduction, e.g. the reproduction of audio and / or video data, and in particular an intelligent and improved way of reproducing information from a plurality of information, such as e.g. a plurality of messages, pieces of music or video clips, but the present invention also relates to the reproduction of information of the type that occurs in home replacement systems, that is to say automatic ordering systems for the household, namely reordering of food.

Especially in the Internet area, there are different methods to present to the Internet user from the wealth of information available on the Internet, either on request or unsolicited, that he is most likely to be interested in due to his personality, i.e. methods that personalize one Offer content. Many of these methods either deliver content on demand, i.e. according to exact specifications, or content according to a set detailed profile.

Other methods, in turn, divide their resources among different users and establish a correlation between the selection of each user. In addition to the explicitly requested content, object recommendations can therefore be offered by other users who have requested similar content.

The disadvantages of the existing methods are on the one hand the need for exact requests or very detailed settings and on the other hand the lack of adaptability to different environments or Situations. In addition, the user must be well acquainted with the respective system in order to be able to enter complex settings.

There is therefore a need for an improved scheme for information reproduction or selection, which is able to reproduce from a large number of items of information, to match the person's personality, the information which is best matched to the personality profile, without the user doing so or to demand complex settings from the person. The ideal would be an information reproduction or selection scheme which, without the user noticing, always selects from the abundance of information available that comes closest to the wishes and needs or, in general, the individualities of the user.

The object of the present invention is therefore to provide an information reproduction scheme which requires little or no additional effort by the user in order to adapt the information reproduction to the user.

This object is achieved by a device according to claim 1 and a method according to claim 24.

The present invention is based on the finding that an intelligent information reproduction scheme which is adaptable to the wishes and needs of the user in a manner which is almost imperceptible to the user and which can be adapted to the user with little or no additional effort can be created, if from information or a plurality of information, the selection of the information to be reproduced is carried out randomly based on an adaptive probability distribution or and randomly and subsequently based on an adaptive probability distribution, and based on a detected behavior of the user the reproduction of the selected information, the adaptive probability distribution is adapted. The behavior of the user is consequently used as a control variable for adapting the adaptive probability distribution, the latter defining, among the reproducible information or the information objects in which the reproducible information is provided, how likely it is that certain information provided in an information object the next random selection for the next information to be displayed or not.

The additional effort for the user in the form of a judgmental behavior to provide the controlled variable for the adaptation of the probability distribution can be kept low or is completely eliminated. According to a special exemplary embodiment, an actuating device is provided for the user in order to be able to express that the current information is to be stopped and new information is to be reproduced, ie a kind of SKIP or skip key. An actuation of this device, which according to the special exemplary embodiment is designed as a simple button, is rated as a negative behavior by the user, while the non-actuation of this device during the playback of certain information is rated as an approving or welcoming behavior of the user. This avoids additional effort for the user, since the user would have operated an actuating device anyway in order to select a next information object himself. On the contrary, ideally, the adaptation of the probability distribution depending on the detected rejecting and approving behavior will result in the user hardly ever having to react negatively by actuating the actuating device. In the event of a negative behavior by the user, the probability distribution is changed such that the reproduction of the information of the information object that has just been reproduced. is less likely next time and at the same time less likely to reproduce information from information objects that more or less fall into the same category as the information object just reproduced. For this purpose, each information object comprises, for example, category assignment data which assign the information object to one or more categories from a predetermined set of categories. The probability distribution is then defined, for example, by a weighting rule that assigns at least one weighting value to each category, which indicates how likely information objects that are assigned to this category will be selected the next time.

According to a special exemplary embodiment of the present invention, the category assignment data is in turn formed by a set of category weights which assign the respective information object to the individual categories with a respective category weighting. For example, a live recording from a musical can be classified as entertainment and music, possibly to different degrees or with different category weightings. A negative behavior of the user therefore has the greatest effect in this exemplary embodiment for those information objects, in addition to the information object, for the information of which the user has shown the negative behavior, which are assigned to the same or the same categories or more precisely with the largest Category weights are assigned to the same or the same category weights.

Further preferred exemplary embodiments of the present invention are explained in more detail below with reference to the accompanying drawings. Show it: 1 is a block diagram of a device for information reproduction according to an embodiment of the present invention; and

Fig. 2 is a schematic diagram illustrating an operation of the device of Fig. 1 according to a specific embodiment of the present invention.

1 shows the basic structure of a device for information reproduction according to an embodiment of the present invention.

As can be seen, the information reproduction device of FIG. 1, which is generally indicated at 10, comprises a control device 12, a reproduction device 14, a device 16 for detecting a behavior of the user of the device 10 in response to information reproduction by means of the reproduction device 14 , a memory 18 for storing a plurality of information objects, in each of which a reproducible information is provided, and a memory 20, in which data are stored, which the control device 12 needs to adapt or adapt the information selection to the personality of the user, such as eg a probability distribution, as will be discussed in more detail below.

After the structure of the device 10 has been described above with regard to its components, its mode of operation and the interaction of its components with one another are described below.

The information reproducing device 10 of FIG. 1 is provided for, from the plurality of information objects that are provided in the memory 18, successively, intermittently or in any other order, Select information objects independently, ie without direct interaction of the user to select the information objects to be reproduced, and reproduce the information provided in the selected information objects. The selection is taken over by the control device 12, which for this purpose has the possibility of accessing the information objects in the memory 18. The reproduction is carried out by the reproduction device 14, which is controlled by the control device 12 for this purpose.

In order not to carry out the selection processes purely randomly or according to criteria that are not tailored to the personality of the user, a probability distribution is stored in the memory 20, which defines for each information object 18 how likely it is that this information object is selected or selected during the selection is assumed (cf. 106 in FIG. 2). The probability distribution, which is stored in the memory 20, is adaptive, so that the control device is effectively connected to the memory 20 not only for reading but also for writing, in order to be able to adapt this probability distribution as described below. To adapt the probability distribution 20, the control device 12 uses data about the behavior of the user in response to the reproduced information of the selected information objects, which it receives from the device 16.

The type of information provided in the information objects, the reproduction of this information in the reproduction device 14 and the behavior of the user in relation to the reproduction of information can take a wide variety of forms and shapes, depending on the purpose for which the information reproduction device 10 is used. However, in order to better illustrate the description of the functioning of the device 10 from FIG. 1, it is assumed below that it is 1 is one which is intended to reproduce audio data, for example as a radio in a vehicle. In this case, for example, the information reproduced in the information objects consists of audio data, such as MPEG files or the like.

The audio data itself can in turn be extremely different from information object to information object and can relate, for example, to news, pieces of music, advertising, glosses, comedy, traffic news, radio plays or weather news. In order to be able to roughly divide or subdivide the abundance of information objects and thus to be able to group information objects into groups, each information object includes not only its actual information but also category assignment data, which each information object either gradually, namely by weighting, or absolutely, by yes or no , one or more of a predetermined set of categories. Possible categorizations include, for example, "music", "classical", "news", "tempo", "energy" etc. and / or the above-mentioned names of different types of audio data, as used in the previous, to show diversity and to represent the diversity of the different audio data. The set of categories can be expandable or fixed.

To give an example, a piece of music from the Beatles can be assigned to the categories music, tempo and energy. The individual assignments could in turn be weighted. For example, a piece of music by the Beatles would be fully, ie 100%, assigned to the category "music" by the category assignment data, while it would only be assigned to the categories "tempo" and "energy" to a small extent. In the case of an audio data reproducing information reproducing device 10, the reproducing device 14 is formed, for example, by a suitable loudspeaker, such as, for example, a car loudspeaker or headphones. In this case, the information or audio data provided in the information objects is reproduced directly to the user, ie the addressee and the location of the reproduction result are at the same location, namely at the user. As is shown with reference to a further application example of the device 10, this need not necessarily be the case.

According to the audio data embodiment, the probability distribution in the memory 20 is now defined by a weighting assignment rule which assigns at least one weighting value to each group of information objects or to each category defined by belonging to a specific category. The controller 12 uses these weighting values when randomly selecting the next information object to be reproduced from the information objects in the memory 18 by more likely selecting information objects that are assigned to a category or categories whose weighting value is greater in the memory 20 according to the weighting assignment rule or weighting assignment table. To determine the probability distribution, as defined by the weighting values of the weighting assignment rule in the memory 20, the control device can furthermore, if present, use the category weights mentioned above, which gradually assign the information objects to the individual categories more or less.

Furthermore, the control device 12 can take into account that some categories are wider and some categories are narrower. For example, the "Music" category applies to more information objects than the "Rock" category because it is a subset of the Music category. insofar however, the weighting value for the category "rock" also applies better to the information objects belonging to it than the weighting value for the weighting category music, and this can then be taken into account by the control device 12 by means of category size weighting values which are assigned to the individual categories, and likewise is used in the selection or definition of the probability distribution among the information objects.

The selection process by the control device 12, which, as described above, is carried out randomly on the basis of the probability distribution in the memory 20 by the control device 12, can be carried out on request by the user, when the end of an audio file being played back or controlled by another system, such as a traffic density monitoring device that only allows audio playback when traffic density is low.

If the control device 12 has made a selection, it outputs the audio information of the selected information object to the playback device 14, which in turn reproduces the audio information to the user.

If the probability distribution on which the selection process by the control device 12 is based, by means of which the information objects are selected, the audio contents of which are output or reproduced in succession or intermittently to the user, would remain unchanged, there would be a great risk that the user would be must be annoyed that he is "bothered" with audio information which, according to his personality, he does not want to hear at all. In other words, the user rejects some of the reproduced information while at the same time finding others well or welcoming them or their reproduction In other words, the user of the device 10 evaluates each displayed information tion or the corresponding information object with an external evaluation, which could be described as "good", "bad" or "ittel". In order to make the internal evaluation of the user detectable for the device 10 and to be able to use it as an adaptation feedback quantity the device 16 the user's behavior to the information reproducing down. This external behavior of the user can then be evaluated or interpreted, namely as a negative attitude of the user, which means that the user did not like the reproduced _information. or approving behavior, which means that the user has welcomed the information that has been reproduced or has enjoyed it.

In the simplest case, the device 16 can consist of a button that the user can actuate to indicate to the device 10 that he does not like the reproduced information. The actuation could then be used at the same time by the control device 12 in such a way that, in response to this, it makes a new selection of an information object based on an adapted probability distribution, as will be discussed below. A device other than a button could of course also be used as the actuating device, e.g. a switch, a lever, a voice input or the like. In any case, the control device 12 would be able to interpret a certain actuation as a negative behavior of the user for the current playback, possibly even with a gradual weighting. For example, the control device could increase the length of time until which the user presses the button mentioned above since the current audio information has been played back, in order to infer how much the user "disapproves" of the audio information currently being played back.

If the control device 12 has determined a negative behavior, it tries the probability model to adapt in the memory 20 such that in the future the selection process will be better adapted to the taste and preferences of the user. In response to detecting a rejection behavior of the user, the controller 12 therefore reduces the weight value for that category for each category associated with the information object to which the user rejected, in the case of using category weights in the Category assignment data of the information object just reproduced with a degree that depends on the category weighting with which this information object is assigned to the individual category. In this way, for example, information objects of these categories will be less likely to be selected next time, since the weighting values for these categories have been reduced compared to the weighting values for other categories.

Although it is possible to record only negative behavior of the user, it is also possible, additionally or alternatively, to use consenting behavior of the user in the adaptation of the probability distribution in the memory 20. Referring again to the previous embodiment for the device 16, namely the button as device 16, the control device 12 could, for example, consider the absence of a button press during the playback of certain audio information as an approving behavior, i.e. in such a way that the user can read the audio information during it Playback did not take place at the push of a button, welcomed or liked it. In the event of an approving behavior, the control device 12 then proceeds in the opposite way to the previous case: it increases the weighting values of the categories to which the information object is assigned, to which the consenting behavior of the user was related, possibly taking into account the category weights of the current one information object. Another option would be to set up two skip buttons 16 use. By pressing one of the skip buttons, for example the right skip button, the playback of the current object would be interrupted, which would be regarded as negative behavior for the current information object. Pressing the other skip key, for example the left skip key, would interrupt the playback of the current object and repeat the playback of the object reproduced in front of the current object, which is an approving behavior for the one reproduced in front of the current object Information object would be evaluated.

The previous example for the device 16, namely a button for recording the behavior of the user in response to the information reproduction, has shown that both an active reaction by the user to information reproduction and the failure to respond to the information reproduction as a negative or consenting behavior can be assessed. Although in the previous the failure to do so was interpreted as consenting behavior, while the active response was interpreted as rejection, a reverse or mixed approach is also possible. For example, the audio data reproducing device 10 as a radio in a car could also be provided with a button which is intended to be pressed by the user in the event that he wishes to express that he would like to hear the audio file which has just been reproduced. Pressing this button several times could then be used for a gradual evaluation in different stages. In the case of voice input to determine behavior, ratings could be entered verbally by the user, e.g. with "good", "bad" etc.

The information reproducing device 10 of FIG. 1 was described in the foregoing against the background of the reproduction of audio data. In the following, the information reproducing device 10 of FIG. 1 is described against the background that it is part of a home Replenishment system is used, namely for example to carry out the reordering of food that has run out or is nearing the end. In this case, the information objects include information that contains specific instructions, namely instructions about the procurement or reordering of certain foods. The control device 12 passes on the information of an information object selected in the selection process by the control device 12 to the playback device 14, which in this case is designed, for example, as a modem, internet connection, fax or another communication device and forwards the order defined by the information to a dealer , which in turn sends the food ordered according to the order to the address of the user of the device 10 or of the home replenishment system, for example by post or another delivery service. All the food delivered to the user and ordered through the home replenishment system ultimately ends up in the user's refrigerator (not shown). In the case of using the information reproducing device 10 for ordering groceries, the place of the reproduction recipient, namely the retailer, and the place of the reproduction result, namely the refrigerator or the user, therefore differ. The connection between the two locations takes place via the delivery service and the order connection.

According to the home replenishment system exemplary embodiment, the information reproduction device 10 now ensures that the refrigerator of the user is constantly refilled. In order for the foodstuffs that are present in the refrigerator to be adapted to the taste of the user, the control device 12 uses to select from among the information objects in the memory 18 at random using the probability distribution in the memory 20 and adapts the probability distribution 20 on the basis user behavior as described below. The orders provided in the information objects in the store 18 relate to those orders of different foods that can be divided into certain categories, such as categories "fruit", "sweet", "sour", "fatty", "light", "Indian", "vegetables", "Asian", " Suitable for diabetics "," vegetarian "etc.

Each information object is now more or less assigned to one or more of these categories. A weighting value is now assigned to each category in the memory 20, with all weighting values defining the probability distribution among the information objects in the memory 18, optionally together with the category weights of the category assignment data in the information objects and the weights mentioned above taking into account the category size.

The device 16 for detecting the behavior of the user in relation to the information reproductions comprises, for example, removing the food from the refrigerator. For this purpose, the refrigerator comprises, for example, a bar code reader as device 16, at which the user guides a foodstuff in order to indicate the removal of this foodstuff from the refrigerator. The device 16 can also be equipped as a camera with object detection. In any case, the control device 12 is always aware of which foods are in the refrigerator, i.e. which ones have just been removed and which ones are being put in or put back in the refrigerator.

The behavior of the user, which he shows in response to the orders, and how it is detected by the device 16, is used by the control device 12 to change the probability distribution in the memory 20 or the weighting values. For example, the control device 12 concludes from the fact that fruit always remains in the refrigerator and it becomes bad that the weighting values for the categories “fruit”, “citrus” fruit "etc. is to be reduced, so that the probability distribution is adapted in such a way that fruit and the like are less likely to be ordered in subsequent orders. Conversely, the control device 12 can, for example, deduce from this that the user delivers yogurt to the refrigerator every time it is delivered will immediately remove this from the refrigerator, conclude that the user has consented, and therefore increase the weighting for the category "yogurt" in response to this, but also the weighting for the category "dairy products" or the like in the memory 20.

From the last exemplary embodiment with reference to FIG. 1, it became clear that the device 16 can be designed in extremely different ways. In general, it can be said that the device 16 should be able to convert the behavior of the user into signals, from which the control device 12 can conclude that the user has consented and / or rejected the reproduced information. For this purpose, the device 16 should, as described above, be able to be sensitive to certain reactions or omissions of the user to the information reproduction, in order to respond to a rejected or rejected information objects when detecting reactions or omissions of this kind. to be able to infer consenting behavior.

In the preceding description, it was assumed that only one weighting value was stored in the memory 20 for each category. In the end, this means that in the preceding functional description of the device of FIG. 1 it was assumed that only an adaptive probability distribution or an adaptive table of weighting values is used to adapt the information object selection. For many applications, this procedure may be a satisfactory result, namely when the preferences of the user are always the same regardless of the situation in which the user expresses his rejection or consent through his behavior. However, this is not always the case. In the previous application example, in which the device of FIG. 1 has been used as an audio information reproducing device, it is possible and also likely, for example, that a user has different preferences or wishes in the morning, for example sitting in the car on the way to work, than in the afternoon when he comes home from work and is tired. In this case, the adaptation result with only one weighting value per category or with only one adaptive probability distribution would not be sufficient, since the probability distribution cannot adapt to fixed preferences or desires of the user, which are not the same in the mornings and afternoons. Therefore, in the following exemplary embodiment, for a functioning of the device according to FIG. 1, in the end, several weighting values are provided for each category or several adaptive probability distributions are provided, which are provided for different situations of the user. A situation-dependent parameter, which is suitable for describing the situation in which the user is currently, is, for example, the time in the previous audio reproduction example. In the home replenishment system example, as described above, it makes sense, for example, to differentiate between "summer" and "winter", since it is quite possible and likely that the user's nutritional preferences will change in summer distinguish between winter and winter, as it is often the case that it is easier to eat in summer, for example more salad, while in winter you eat more heartily, for example more roasts. For example, eating habits can also differ between working days and weekend days. Other situation-related parameters can be the season, the day of the week and additional environmental parameters that are relevant for the respective application.

The operation of the device of FIG. 1 is therefore described in more detail with reference to FIG. 2 with reference to a further exemplary embodiment, in which the device acts as an audio playback device. According to this exemplary embodiment, each information object 50 among the information objects that are stored in the memory 18, in addition to information data 52 in which the actual information of the information object 50 is provided, comprises object-specific parameters 54 that correspond to the previous category assignment data that contain the actual information in FIG describe the information data 52, and which also have the category weights 56 mentioned above, each of which is assigned to a specific one of the categories 58 and for which indicates how much the information object 50 is assigned to this category 58. The categories listed by way of example in FIG. 2 are music, classical music, news, tempo and energy. The category weights 56 are given as an example in percent, although other representations are also possible.

Global weighting data 60, which are stored in the memory 20, are used to determine the adaptive probability distribution. For each category 58, this global weighting data comprises not only a weighting value, as was described above with reference to FIG. 1, but also a plurality of weighting values which are assigned to different possible values of a situation-related parameter. In order to be able to divide the situations in which the user finds himself more precisely, several situation-related parameters 62 are even provided according to the exemplary embodiment of FIG. 2, and for each of these situation-related parameters 62 the global weighting data 60 in the memory 20 comprise a plurality weighting values per category 58, namely a weighting value per possible quantization value of the respective situation-related parameter. Exemplary examples of situation-related parameters 62 are indicated in FIG. 2 with “time of day”, “mood”, “position” and “weather situation”. For example, let N _κ the number of categories 58, N _SP the number of situation-related parameters 62 and n _S p (l) the number of distinguishable quantization values for the first situation-related parameter, n _S p (2) the number of quantizations for the second situation-related parameter Parameters, etc., then the global weighting data 60 would be a number of

"5f, ∑ n _SP (i) i = l

Weighting values.

Instead of the tabular storage of the global weighting data 60, it would also be possible, as indicated in FIG. 2, for an analytical function course, which is dependent on the respective situation-related parameter t for the time of day, to be stored for each tuple from category 58 and situation-related parameter 62. s for the mood, p for the position or w for the weather to a weighting value g.

Now that the differences with regard to the definition of the probability distribution by weighting values in the memory 20 from the exemplary embodiments described with reference to FIG. 1 have been discussed above with reference to FIG. 2, the mode of operation of the device from FIG 1 described in the context of the embodiment of FIG. In the exemplary embodiment in FIG. 2 too, the device 10 is provided for successively storing information objects from the plurality of information objects in the memory 18 with the aid of an adapted probability distribution. tion as defined by the global weighting 60, to be selected randomly and to adapt the global weighting data 60 based on the user behavior to the reproduction of the information in the selected information objects.

In contrast to the exemplary embodiments which have been described directly with reference to FIG. 1, the device 10 according to the exemplary embodiment of FIG. 2, together with the user behavior, detects information reproduction as captured in a user behavior detection process 64, however also the instantaneous situation-related parameters in a situation detection process 66. The situation-related parameters are detected by a situation parameter detection device 68, which is shown in FIG. 1 with dashed lines. The result of the situation detection process 66 are recorded values for the situation-related parameters, here the current time of day, the current mood of the user, the current position of the user and the current weather situation. The device 68 may for this purpose a clock for detecting the time of day, a voice analyzer with pattern recognition, a ^'camera device, a heart rate sensor on the steering wheel, a voice recorder, or a device for evaluation of behavioral patterns in general for identifying the mood of the user, a GPS sensor to detect the position of the user and have a combined brightness, moisture, air pressure and wind speed sensor to record the weather. The current values of the situation-related parameters are indicated at 70 in FIG. 2.

As soon as the behavior detection 64 detects a rejecting or consenting behavior of the user, which indeed represents an "external evaluation" of the current selection among the information objects 50, this external evaluation goes together with the associated current values 70 of the situation-related parameters in an evaluation process 72. The evaluation process 72 takes over the adaptation of the weighting values already described above with reference to FIG. 1 in the context of the previous exemplary embodiments. The adaptation of the weighting values on the basis of the external evaluation and the current values 70 of the situation-related parameters is illustrated by an active arrow 74. The adaptation 74 is now carried out in the following manner. If the behavior detection 64 yields a negative external evaluation or a negative attitude to the information object 50, the information data 52 of which is currently being reproduced, the evaluation process 72 uses the object-specific parameters 54 of this current information object 50 and the current values 70 of the situation-related parameters to adapt the global weighting table 60.

In particular, in the event of a negative behavior by the user, the evaluation process 72 acts on the global weighting table 60 such that subsequently information objects with object-specific parameters 54 similar to the current information object 50 are selected less likely, but only to the extent that the situation is the same or similar is, ie for all cases in which the situation-related parameters have values similar to the current values 70, or only for those weighting values in the memory 20, the situation parameter values are assigned which are the same or similar to the current 70. Still more specific Expressed reduced the evaluation process 72 under all weight values, which are one and the same category 58 and one and the same situation-related parameters 62 associated with those at the ^'most, which are the current value of the situation-related parameter at the next. In other words, the impact of the adaptation due to the negative behavior of the user decreases with increasing deviation of the situation, to which the respective weighting value relates, from the current situation. tion 70 from. This takes into account the fact that a user who, for example, made a negative decision for a rock piece of music in the morning, for example at 8:00 a.m., would probably have made this decision at 9:00 a.m. and maybe also at 10:00 a.m. ; but for example at noon, during the lunch break, he may want to hear rock again. The rejecting behavior of the user therefore radiates something in the case of the adaptation 74 in order to also have an effect on weighting values which relate to similar situations. In this way, the adaptation can be accelerated.

Within the weighting values that relate to the same situation, that is to say the same possible value of a respective situation-related parameter 62, the evaluation process 72 most reduces those weighting values that relate to a category that the current information object 50 has the most due to its object-specific parameters 54 is assigned, ie its category weighting 56 is greatest in the object-specific parameters 54 of this object 50. This takes into account the fact that a negative attitude towards a rock piece of music should not lead to news being played less often, but primarily that rock pieces and similarly energetic pieces of music should no longer be played as likely. Another example would be the rejection of a piece of music that can be 100% assigned to pop, but also has a 20% jazz influence. In this example, the weighting only had a 20% effect on the global jazz rating. In the adaptation 74, as described above, weights may also play a role, which are firmly assigned to the categories 58, in order to take into account that some categories relate to a large number of objects, while some categories only relate to a smaller group can relate to objects because they are pure subgroups of the former and therefore more specific. In the case of a positive external evaluation or an approving behavior of the user, the evaluation process 72 in the case of the adaptation 74 behaves in exactly the opposite way to the previous description, namely in that the values are increased instead of decreased, and in the same way, namely all the more , the closer the value of the situation-related parameter to which the respective weighting value comes to the current value 70, and the more, the greater the category weighting 56 for the category to which the respective weighting value relates in the object-specific Parameters 54 of the object 50, to which the consenting behavior relates.

In the further description of FIG. 2, it is assumed that a behavior detection device 16 is used for behavior detection 64, which is designed as a button that is used by the user to abort the reproduction of a current information object and thus to express a negative behavior It is rated as positive behavior if the user does not press the button while playing an information object.

If the user now presses the button of the behavior detection device 16, the evaluation process 72 enters this as part of an update 76 in an object-specific weighting table 78, which is stored in the memory 20 together with the global weighting table 60, for example. The evaluation process 72 stores negative behavior of the user in the object-specific weighting table 78. In particular, when the user rejects behavior, the evaluation process 72 enters an index for the information object 50, to which the rejection behavior referred, together with the current values 70 of the situation-related parameters on which the rejection behavior took place, into the object-specific weighting table 78. The logging of the negative behavior serves to adapt the information more quickly. mation object selection process, as will be discussed in more detail below. Table 78 is limited in scope. The evaluation process 72 therefore displaces an old entry of an earlier rejecting behavior when a current rejecting behavior of the user is entered by means of a suitable displacement strategy, such as a FIFO (first in first out = first in, first out) strategy or the like. The scope of table 78 can include, for example, 1,000 entries to be rejected.

As indicated above, it is assumed below that the user in the exemplary embodiment of FIG. 2 expresses a rejection by pressing a button. The push of a button then causes the updates 74 and 76 mentioned above. In addition, the push of a button triggers the next selection process 80, since the user not only expresses a negative behavior by pushing a button, but also informs the system that the current playback stop and pause and play a new information object. The triggering at the push of a button is indicated in FIG. 2 with the arrow 82. The selection process 80 of FIG. 2 is a two-stage selection process. The first stage, indicated at 84 in FIG. 2, corresponds to the selection process already described with reference to FIG. 1 on the basis of the global weighting 60. The second stage, indicated at 86 in FIG. 2, is based on the object-specific weighting 78.

The first stage 84 of the selection process 80 is carried out by the control device 12 in the following manner according to the exemplary embodiment of FIG. 2. First, the control device 12 uses the global weighting data (88) in order to calculate in a calculation process 90 those object-specific parameters to which the highest probability is assigned according to the global weighting data 60. In other words, the control device calculates 12 in the calculation process 90, a set of category weights 56 in such a way that the highest selection probability is assigned to this set of the probability distribution defined by the global weighting data 60, specifically with reference to the current values 70 of the situation-related parameters. If there really were such an object with such object-specific parameters among the information objects in the memory 18, then this would most likely be accepted, taking into account the current situation. In another situation, of course, the calculation process 90 would lead to a different set given the same global weighting data. In other words, the resulting set of category weights shows a situation-related rejection / acceptance level of each category assignment.

Regardless of this, the control device 12 makes a random selection from among the information objects in the memory 18 in a random process 92. The random process 92 selects each information object with the same probability, for example. However, a different probability distribution could also be provided for step 92, which, however, is fixed in contrast to the adaptive probability distribution.

The two processes 90 and 92 lead to two sets of category weights, namely a calculated optimal object-specific parameter set or an optimal set 94 of category weights from the calculation process 90, which, as described above, is situation-dependent, and a set 96 of category weights that the corresponds to object-specific parameters 54 of the information object 50 selected in the random process 92. The two sets 94 and 96 are the input data for a parameter comparison 98, which the control device 12 carries out in order to complete the first stage 84 of the selection process 80. In particular, in the parameter comparison 98, the two sets 94 and 96 are used, for example, by scalar multi- the two category weighting vectors 94 and 96 a probability value is formed and used in a random process 100 in order to randomly accept or reject the object selected in the random process 92 based on this probability value. The probability of the acceptance of the information object selected in the random process 92 is higher, the more the object-specific parameters 96 of the selected object correspond to the optimal object-specific parameters 94.

In the event that the random process 100 of the parameter comparison 98 leads to a rejection, the processes 90 and 92 are carried out again, as is indicated by an arrow 102 marked "New selection due to rejection". In the case of the acceptance of the im Step 92 of a randomly selected object in the parameter comparison 98 feeds this information object to a further acceptance / rejection stage, namely the second stage 86 of the selection process 80, this sequence being indicated in FIG. 2 with an arrow headed “acceptance”.

The stage 86 comprises a validity check process 104, which is based, among other things, on the object-specific weighting data 78, as is indicated by an arrow 106. Validation 104 is either random and / or absolute in nature. For example, the validity check 104 checks whether there is an entry in the data 78 which relates to the information object selected in the first stage 84 of the selection process 80, and whether this entry has values for the situation-related parameters 62 which differ from do not differentiate the current values 70 by more than a predetermined amount. This would mean that, in the past, the user has rejected exactly this information object in a similar situation. In the case of a deterministic validity check 104, this could lead to rejection. In the case of a weighted random Depending on the distinction between the current values 70 and the values for the situation-related parameters in the entry in table 78, the rejection 104 could vary in probability. In this way, the validity check acts in such a way that the slower adaptation 74 is preceded by a faster adaptation 76.

The validity check 104 can also use additional criteria for the validity check. For example, the evaluation process 72 logs the reproduced information objects in a further list (not shown in FIG. 2). An immediate repetition of these information objects in a period of, for example, two hours could then be avoided in the validation process 104, for example, by rejecting the object that passed the stage 84 in the stage 86. Furthermore, the validation process 104 could also use data in the object currently to be checked itself in order to make a rejection or acceptance decision. For example, in the case of message information objects, the age of the message or the timeliness of the message could be included in the process 104. Older messages would be less likely to go through the validation process 104 with an acceptance.

If the validation process 104 leads to a “rejection”, the selection process 80 starts again at the processes 90 and 92, as indicated by the arrow 102. With each rejection, the strictness of the review of the rejection criteria for the subsequent run decreases to ensure that a selection is made after a maximum number of runs. However, if the object has also passed the second stage 86 by "acceptance", then this object is the object selected in the selection process and is output as an object selection result to the reproduction device 14 as it is is indicated by an arrow 108. From the reproduction of the object selection, the behavior detection 64 now relates to this selected object, as indicated by an arrow 110. This closes the feedback loop for adaptation 74 and 76, which includes the user's assessment as the basis for the adaptation.

In other words, the device of FIGS. 1 and 2 effects a personalized selection of information. The behavior of the device is influenced in the exemplary embodiment of FIG. 2 on the basis of the consideration of the situation and environmental reaction, with information selection being understood to mean the selective storage and reproduction of information. The procedure corresponds to a feedback system that approaches the desired selection behavior after the environmental reaction to the previous selection of information. The environmental reaction is specified on the basis of several input variables, namely the various environmental parameters or situation-related parameters and the selection evaluations or the external evaluation. The evaluation is generated on the basis of the objects already selected by an external evaluation device, for example the user.

The information to be selected is structured and stored in the form of independent objects. In the above exemplary embodiments, an object consisted of pure information data and of object-specific parameters that described the object content. The environment to which the system responded was described with situation-related parameters.

If, for example, the device of FIG. 1 is initialized according to the exemplary embodiment of FIG. 2, an object is initially selected from the set of available objects by an equally distributed stochastic random process. For subsequent stochastic selection processes, the selection probabilities 60 are adjusted depending on the external evaluations to date and the current situation. Or more precisely and with regard to that Expressed in the exemplary embodiment of FIG. 2, the selection from the memory is always made at random, although the probability of acceptance is continuously adapted. The influence of an external evaluation or a user decision on the current selection process or subsequent selection processes under similar situations is weaker the more the current situation deviates from the situation in which the evaluation was given or carried out, which here as degressive procedure is called.

The above exemplary embodiments enable the personalized selection of information taking into account external environmental reactions without detailed specification of selection criteria. Complex active user involvement is avoided in these exemplary embodiments.

A possible application example of a device according to FIG. 1, which operates according to FIG. 2, is the selection of audio objects according to the preferences of a user in a car radio. In addition to the pure information data 52, such as pieces of music, an audio object 50 contains object-specific parameters 54 that describe the object 50, such as object type, ie music, news, advertising etc., object length, object category, ie rock, classical, techno or Sports, business, abroad, etc. Possible situation-related parameters 62 are the time of day, the mood of the user, the location, date / season, the weather situation, etc. As described above, the evaluation option can be a “Skip” button With which the user can skip the object, the audio objects 50 are offered to the user, with the automatic selection approaching the preferences of the user over time based on the external evaluations Listens to news or prefers blues when it rains. In addition to this application example, there are many other application options for personalized information selection in accordance with the previous selection examples. These include, for example, the selection of the daily clothing combination, the adapting color scheme or the scent of a room, the selection of food in the refrigerator in a so-called home replenishment system, as shown above, etc.

With reference to the preceding description, it is pointed out that the memory 18 in which the information objects are provided can be, for example, a CD, a hard disk, a DVD, a magnetic memory or another memory in which a fixed set of information objects is stored is. The set of information objects could then be changed intermittently by changing the CD or updating the set. However, it is also possible for the memory 18 to be designed as a cache, in which information objects are stored, which are broadcast cyclically, for example, via a radio signal. The cyclically broadcast information objects are received at a radio receiver and then entered in the cache memory according to a predetermined displacement strategy or displaced again from the memory 18. The displacement strategy could use the probabilities defined by the global weighting 60, the object-specific weighting data 78 and the further criteria used by the validity check 104. An information object, which is selected very unlikely in the selection process 80 according to the criteria mentioned, is therefore not transferred to the cache memory 18 at all or is displaced again very early. The displacement strategy could also use the data 60, 78 and the further data used by the validity check 104 as well as the usual cache criteria for displacement, such as according to the FIFO principle, ie it could be a modified FIFO principle or the like. Furthermore, it is pointed out that, in contrast to the previous exemplary embodiments, the adaptation of the probability distribution is not restricted to the adaptation to the needs and wishes of the user. It could also be that, in the case of a car radio, for example, the information selection is not adapted to the probability distribution by the user pressing a button in order to abort an information object and request a new one, but that, for example, the driving behavior of the user is used to increase the probability distribution adapt. If the aggressiveness of the driving behavior of the user increases significantly, for example, when playing a rock piece of music, this is taken into account by adapting the probability distribution 60, in that rock pieces of music and similar audio objects are less likely to be reproduced in similar situations. Such an adaptation could increase driving safety, for example in the case of trucks.

With regard to the evaluation or the evaluation process 72, it is also pointed out that the evaluation of the reduction in the underlying parameters described there could also only take place if parameters have to be calculated for a similar situation. In other words, the degression process could only take place after selection instead of evaluation.

The control device 12 mentioned above could, for example, be a computer or an appropriately designed ASIC.

In particular, it is pointed out that, depending on the circumstances, the scheme according to the invention can also be implemented in software. The implementation can be carried out on a digital storage medium, in particular a floppy disk or a CD with electronically readable control signals, which can thus be carried out with a programmable computer system can cooperate that the corresponding procedure is carried out. In general, the invention thus also consists in a computer program product with program code stored on a machine-readable carrier for carrying out the method according to the invention when the computer program product runs on a computer. In other words, the invention can thus be implemented as a computer program with a program code for carrying out the method if the computer program runs on a computer.

Claims

claims

Apparatus for reproducing information provided in information objects (50), comprising means (12) for randomly selecting an information object (50) based on an adaptive probability distribution (60) in order to obtain a selected information object; means (14) for displaying information provided in the selected information object; means (16) for detecting a user's behavior upon reproduction of the information provided in the selected information object; and means (12) for adapting the adaptive probability distribution based on the detected behavior.

2. Apparatus according to claim 1, wherein each information object (50) is assigned category assignment data (54) which assign the same to one or more categories from a set of categories, the probability distribution being at least partially defined by a weighting assignment rule which is assigned to each category (58) assigns at least one weighting value, and the device (12) is designed for random selection in order to carry out the selection depending on the weighting assignment rule and the category assignment data which are assigned to the selected information object.

3. Apparatus according to claim 1 or 2, wherein each information object (50) has a set of category weights. assignments (56) are assigned as category assignment data (54), of which each category weighting (56) is in turn assigned to a category (58) from a set of categories, and the probability distribution is at least partially defined by a weighting assignment rule (60) which assigns at least one weighting value to each category (58), and the device (12) is designed for random selection in order to carry out the selection depending on the weighting assignment rule and the category weights assigned to the selected information object.

4. The device according to claim 2 or 3, further comprising means (68) for detecting a situation-related parameter, which describes a situation in which the detected behavior of the user takes place in order to obtain a detected value (70) for the situation-related parameter , wherein the weighting assignment rule assigns a weighting value to each category for different possible values of the situation-related parameter, the device (12) being designed for random selection in order to make the selection depending on the weighting assignment rule, the detected value of the situation-related parameter and the category assignment data corresponding to the assigned to the selected object.

5. The device of claim 4, wherein the situational parameter is a current time of day, a season, a day of the week, a current mood of the user, a current position of the user, or a current weather situation.

The apparatus of any one of claims 2 to 5, wherein the random selection means (12) further comprises: means (90) for determining, based on the adaptive probability distribution, optimal category assignment data (94); means (92) for randomly selecting a candidate information object from the information objects; means (98) for accepting the candidate information object as the selected information object in a random manner depending on a first random decision (100) with a first probability which depends on the optimal category assignment data (94) and the category assignment data (96) assigned to the candidate information object, wherein the means for random selection and the means for accepting are designed to repeat the selection and acceptance if the candidate information object is rejected.

7. The device according to claim 6, wherein the means (98) is designed to accept, in order to further carry out the candidate information object depending on a second decision (104) with a second probability, which depends on additional data that is dependent on the weighting assignment rule and differ from the category assignment data and only accept the candidate information object if both random decisions are positive.

8. The device of claim 7, further comprising: means (78) for managing a list, in which the negative behavior of the user for the reproduction of information objects or reproduced information objects is stored, the means for accepting being designed to use the list at least as part of the additional data.

9. The apparatus of claim 8, wherein the additional data has an age of the information of the candidate information object.

10. The device according to one of claims 6 to 9, further comprising a device (68) for detecting a situation-related parameter, which describes a situation in which the detected behavior of the user takes place, by a detected value (70) for the situation-related parameter to be obtained, the weighting assignment rule assigning a weighting value to each category for different possible values of the situation-related parameter, the device being designed to determine the optimal category assignment data in order to carry out the determination based on the weighting assignment rule and the detected value of the situation-related parameter in order to obtain a situation-dependent optimal set of category weights as optimal category assignment data.

11. The device according to one of the preceding claims, further comprising: a memory (18) in which the information objects are stored.

12. The apparatus of claim 11, wherein the memory (18) comprises a removable data carrier, a CD, a DVD, a hard disk or a magnetic memory.

13. The apparatus of claim 11, further comprising: means for receiving information objects that are broadcast over a broadcast signal; and means for replacing information objects in the memory with received information objects according to a predetermined displacement strategy.

14. The apparatus of claim 13, wherein the displacement strategy depends on the adaptive probability distribution.

15. Device according to one of the preceding claims, wherein the device (16) for detecting the behavior of the user comprises an actuating device for entering a reaction of the user.

16. The apparatus of claim 15, wherein the means for detecting the behavior of the user is designed to interpret the input of a response of the user to the reproduction of the information provided in the selected information object as a negative behavior of the user.

17. The apparatus according to claim 15, in which the device for detecting the behavior of the user is designed in such a way that, in the event that the user does not enter a reaction to the reproduction of the information provided in the selected information object, this as an approving behavior to interpret .

18. The apparatus of claim 15, wherein the device for detecting the behavior of the user has two input options and is designed to input a reaction of the user to the reproduction of the information provided in the selected information object using the first input option as a rejecting behavior to interpret the user and to interpret the input of a reaction of the user to the reproduction of the information provided in the selected information object using the wide input option as the user's consenting behavior.

19. The device according to any one of claims 15 to 18, wherein the input device comprises a button, a microphone, a camera, a lever or a switch.

20. Device according to one of the preceding claims, wherein the information comprises audio and / or video data or reorders for a home replenishment system.

21. Device according to one of claims 2 to 20, wherein the device is adapted to adapt, in the event that the detected behavior is a rejecting behavior, to adapt the adaptive probability distribution in such a way that subsequently information objects belonging to a category are assigned ^', which is associated with the information object, in which the information reproduced is provided, are less likely to be selected by the means for supply due selecting, and / or in the case that the detected behavior is an approving behavior, the adaptive probability distribution to adapt in such a way that information objects, which are assigned to a category to which the information object which has provided the reproduced information is assigned are more likely to be selected by the device for random selection.

22. The apparatus according to claim 21, wherein the device is designed to adapt to carry out the adaptation depending on a current value (70) of a situation-related parameter that describes the situation in which the detected behavior of the user takes place.

23. The apparatus of claim 21 or 22, further comprising means (68) for detecting a situation-related parameter that describes a situation in which the detected behavior of the user takes place in order to obtain a detected value (70) for the situation-related parameter , wherein the weighting assignment rule assigns a weighting value to each category for different possible values of the situation-related parameter, and in which the device is designed to adapt to carry out the adaptation depending on the detected value, such that the weighting values assigned to this category are changed for each category The more the possible value of the situation-related parameter, for which the respective weighting value is assigned to the respective category, deviates from the recorded value, the more changes are made. -

24. A method for reproducing information provided in information objects (50), comprising the following steps: random selection of an information object (50) based on an adaptive probability division (60) to obtain a selected information object;

Reproducing information provided in the selected information object;

Detecting a behavior of a user in response to the reproduction of the information provided in the selected information object; and based on the detected behavior, adapting the adaptive probability distribution.

25. Computer program with a program code for carrying out the method according to claim 24, when the computer program runs on a computer.